because the value function (C.2) associated with this strategy is unbounded at Z = n∕δ and
thus the strategy a(Z)=0for Z ∈ [0, ∞) ceases to be continuous at this point.
Next, consider the case where c2 =0. In this case it turns out that the derivative of the
resulting value function with respect to Z is bounded above:
V0(Z)=
n - δZ
n (ρ + 3δ)
(n - 3δZ) ρ+
< 0 for ∀Z>0.
(C.3)
On other hand, since
lim dpiZ = lim r (n- 1) Z → ∞ for ∀Z > 0, (C.4)
ai→0∂ai ai→0 n2a
inequality (∂pi/dai) Z < V0 (Z) never holds except for Z = 0. Hence, the cornered strategy
α(Z) = 0 is not an equilibrium strategy for Z ∈ (0, ∞). ■
Lemma 2 There exists a constant of integration which makes the strategy a(Z)=1 an equi-
librium strategy over Z ∈ (0, ∞).
Proof. When all players play strategy ai = 1, the HJB equation (6) becomes
ρV (Z) = nZ + V' (Z)(-δZ).
(C.5)
By integration and imposing symmetry, we have
V(Z)=
Z + Z ρn (ρ + δ) C3
n(ρ+δ)
(C.6)
where c3 represents a constant of integration. When setting a = 1 in ∂pi∕∂ai yields (∂pi∕∂ai)Z ≡
r (n — 1) Z/n2, the first-order condition (∂pi∕∂ai) Z ≥ V0 (Z) only allows (C.6) to hold for the
values of Z satisfying
δ 1 .+’ r (n — 1) -, 2'
— ----Z δ--------Z δ
≤ c3.
(C.7)
ρn ρ + δ n
Since the first exponent inside the brackets on the left-hand side of (C.7) is smaller than the
second exponent term, it dominates for smaller values of Z, whereas for larger values of Z
25